During an event held in London, Meta confirmed its upcoming release of Llama 3, the next-generation large language model (LLM) that powers generative AI assistants. This announcement aligns with a recent report by The Information, indicating Meta’s imminent launch.
Nick Clegg, Meta’s president of global affairs, mentioned that the rollout of Llama 3 will commence within the next month or even sooner, with various models featuring different capabilities set to debut throughout the year. Chris Cox, Meta’s Chief Product Officer, added that Llama 3 will be integrated across multiple Meta products.
Meta has been striving to catch up with OpenAI since the viral success of ChatGPT, which created mainstream interest in generative AI. Llama 3, a more extensive model than its predecessors, aims to address previous criticisms of limited functionality by offering improved accuracy and handling a broader range of topics, including potentially controversial ones.
Joelle Pineau, Vice President AI Research at Meta, expressed the company’s goal for Llama-powered AI to become the most useful assistant globally. However, Meta has not disclosed the exact parameters of Llama 3 or provided demos of its functionality yet.
Meta’s approach with Llama, as an open-source project, contrasts with more proprietary models, catering to developer preferences. Despite this, Meta remains cautious, particularly regarding other forms of generative AI like image generation, as it focuses on safety, latency, and usability.
Yann LeCun, Meta’s chief AI scientist, expressed skepticism about the limitations of generative AI and advocated for a different approach called joint embedding predicting architecture (JEPA). This signals Meta’s ongoing exploration of diverse AI strategies beyond generative models.
Source link
